From af4f9aea5a36b46ae13cd2d336316fb91fbd7c35 Mon Sep 17 00:00:00 2001 From: =?utf8?q?J=C3=A9r=C3=B4me=20Benoit?= Date: Tue, 7 Oct 2025 01:11:04 +0200 Subject: [PATCH] docs(reforcexy): improve win_reward_factor documentation MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit Signed-off-by: Jérôme Benoit --- ReforceXY/reward_space_analysis/README.md | 4 ++-- ReforceXY/reward_space_analysis/reward_space_analysis.py | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/ReforceXY/reward_space_analysis/README.md b/ReforceXY/reward_space_analysis/README.md index 683ab27..d657c78 100644 --- a/ReforceXY/reward_space_analysis/README.md +++ b/ReforceXY/reward_space_analysis/README.md @@ -268,7 +268,7 @@ _Efficiency configuration:_ _Profit factor configuration:_ -- `win_reward_factor` (default: 2.0) - Amplification for PnL above target (no upper bound; effective profit_target_factor ∈ [1, 1 + win_reward_factor] because tanh ≤ 1) +- `win_reward_factor` (default: 2.0) - Asymptotic bonus multiplier for PnL above target. Raw `profit_target_factor` ∈ [1, 1 + win_reward_factor] (tanh bounds it); overall amplification may exceed this once multiplied by `efficiency_factor`. - `pnl_factor_beta` (default: 0.5) - Sensitivity of amplification around target _Invariant / safety controls:_ @@ -704,7 +704,7 @@ Before simulation (early in `main()`), `validate_reward_parameters` enforces num | `exit_half_life` | 1e-6 | — | Half-life in duration ratio units | | `efficiency_weight` | 0.0 | 2.0 | Blend weight | | `efficiency_center` | 0.0 | 1.0 | Linear pivot (efficiency ratio center) | -| `win_reward_factor` | 0.0 | — | Amplification for pnl above target | +| `win_reward_factor` | 0.0 | — | Asymptotic bonus multiplier for pnl above target | | `pnl_factor_beta` | 1e-6 | — | Sensitivity ≥ tiny positive | Non-finite inputs are reset to the applicable minimum (or 0.0 if only a maximum is declared) and logged as adjustments. diff --git a/ReforceXY/reward_space_analysis/reward_space_analysis.py b/ReforceXY/reward_space_analysis/reward_space_analysis.py index 17fdc55..cdddc64 100644 --- a/ReforceXY/reward_space_analysis/reward_space_analysis.py +++ b/ReforceXY/reward_space_analysis/reward_space_analysis.py @@ -166,7 +166,7 @@ DEFAULT_MODEL_REWARD_PARAMETERS_HELP: Dict[str, str] = { "exit_half_life": "Half-life for exponential decay exit mode.", "efficiency_weight": "Weight for efficiency factor in exit reward.", "efficiency_center": "Pivot (in [0,1]) for linear efficiency factor; efficiency_ratio above this increases factor, below decreases.", - "win_reward_factor": "Amplification for pnl above target (no hard cap; asymptotic).", + "win_reward_factor": "Asymptotic bonus multiplier for pnl above target: approaches (1 + win_reward_factor); combined with efficiency_factor the final product can exceed this bound.", "pnl_factor_beta": "Sensitivity of amplification around target.", "check_invariants": "Boolean flag (true/false) to enable runtime invariant & safety checks.", "exit_factor_threshold": "If |exit factor| exceeds this threshold, emit warning.", -- 2.43.0